Transmembrane Topology of the Protein Palmitoyl Transferase Akr1*

The two recently identified protein acyl transferases (PATs), Akr1p and Erf2p/Erf4p, point toward the DHHC protein family as a likely PAT family. The DHHC protein family, defined by the novel, zinc finger-like DHHC cysteine-rich domain (DHHC-CRD), is a diverse collection of polytopic membrane proteins extending through all eukaryotes. To define the PAT domains that are oriented to the cytoplasm and are thus available to effect the cytoplasmically limited palmitoyl modification, we have determined the transmembrane topology of the yeast PAT Akr1p. Portions of the yeast protein invertase (Suc2p) were inserted in-frame at 10 different hydrophilic sites within the Akr1 polypeptide. Three of the Akr1-Suc2-Akr1 insertion proteins were found to be extensively glycosylated, indicating that the invertase segment inserted at these Akr1p sites is luminally oriented. The remaining seven insertion proteins were not glycosylated, consistent with a cytoplasmic orientation for these sites. The results support a model in which the Akr1 polypeptide crosses the bilayer six times with the bulk of its hydrophilic domains disposed toward the cytoplasm. Cytoplasmic domains include both the relatively large, ankyrin repeat-containing N-terminal domain and the DHHC-CRD, which maps to a cytosolic loop segment. Functionality of the different Akr1-Suc2-Akr1 proteins also was examined. Insertions at only 4 of the 10 sites were found to disrupt Akr1p function. Interestingly, these four sites all map cytoplasmically, suggesting key roles for these cytoplasmic domains in Akr1 PAT function. Finally, extrapolating from the Akr1p topology, topology models are proposed for other DHHC protein family members.

Many proteins, particularly many signaling proteins, rely on covalent lipid modifications, either prenylation, myristoylation, or palmitoylation, for membrane attachment. Being limited to the cytoplasmic compartment, these lipid modifications serve to tether proteins to the cytoplasmic surface of cellular membranes. The enzymes that mediate myristoylation and prenylation, i.e. the myristoyl and prenyl transferases, are soluble cytoplasmic proteins. In contrast, the palmitoyl transferases, i.e. the protein acyl transferases (PATs), 1 the first examples having been identified only recently, are predicted to be polytopic membrane proteins (1,2). To gain insight into the PAT domains that are available cytoplasmically to mediate this lipid modification, we have determined the transmembrane topology of the prototypical PAT, the yeast protein Akr1p.
The first PATs were identified through recent work in the yeast Saccharomyces cerevisiae. Akr1p was identified as the PAT for the two plasma membrane-localized, type I casein kinases, Yck1p and Yck2p, whereas Erf2p and Erf4p were found to collaborate for the palmitoylation of the yeast Ras proteins, Ras1p and Ras2p (1,2). The two yeast PATs are substrate-specific. Although Ras2p is mislocalized from the plasma membrane in erf2⌬ or in erf4⌬ cells, the plasma membrane localization of Yck2p remains unperturbed. Likewise, in akr1⌬ cells, Yck1p and Yck2p mislocalize, whereas Ras2p plasma membrane localization remains unaffected (1). Akr1p localizes to the Golgi (1, 3), whereas Erf2p and Erf4p localize to the endoplasmic reticulum (4,5). Palmitoylation of Yck1p and Yck2p is fully dependent on Akr1p in vivo, being abolished in an akr1⌬ cell context. Furthermore, when purified to apparent homogeneity, Akr1p alone promotes the in vitro transfer of the palmitoyl moiety from palmitoyl-CoA to the Yck2 substrate protein, indicating that no accessory subunits are required for Akr1 PAT activity (1). In contrast, Erf2p and Erf4p appear to collaborate for Ras2p palmitoylation, both in vivo and in vitro (2,4), suggesting that these two proteins may be co-subunits of the Ras PAT. As the 86-kDa Akr1p is notably larger than either the 41-kDa Erf2p or the 27-kDa Erf4p, Akr1p may represent an evolutionary melding of the two functions into a single polypeptide.
Although the two PATs do not share extensive sequence homology, Akr1p and Erf2p are both predicted to be polytopic membrane proteins, and both contain the 51-residue-long novel zinc finger-like sequence, the DHHC cysteine-rich domain (DHHC-CRD). The DHHC-CRD sequence defines a diverse protein family extending through all eukaryotes; yeast has seven DHHC proteins, whereas 23 have been identified from the human genome. Aside from the DHHC-CRD, DHHC proteins share little sequence homology. Nonetheless, all are predicted to be polytopic membrane proteins, and all have the DHHC-CRD sequence analogously positioned between predicted transmembrane domains (TMDs). Beyond the two PATs, Akr1p and Erf2p, little is known regarding the biochemical function of other DHHC family members. An intriguing possibility is that these proteins may all be PATs. As mutation of conserved DHHC-CRD residues abolishes PAT activity for both Akr1p and Erf2p (1,2), the DHHC-CRD sequence, the single element preserved throughout this family, could represent the conserved elements of the PAT active site. The substantial diversity embodied within this family perhaps accommodates the diverse range of substrate proteins that are known to be palmitoylated (6 -8).
Several mammalian DHHC proteins have been identified in two-hybrid searches. GODZ was identified by its interaction with the glutamate receptor ␣1 subunit (9); HIP14, the human orthologue of Akr1p, was identified through its interaction with huntingtin, the disease protein of Huntington's disease (10); and Aph2 was identified through its interaction with c-Abl tyrosine kinase (11). GODZ and HIP14 have both been localized to the Golgi, whereas Aph2 has been localized to the endoplasmic reticulum (9 -12). Consistent with a general DHHC protein role in protein palmitoylation, a recent study of the rat GODZ indicates a role for this protein in the palmitoylation of the ␥2 subunit of the pentameric GABA A receptor (12).
In addition to its DHHC-CRD, which has been implicated in catalysis, Akr1p also has a domain for protein-protein interaction, a stretch of six ankyrin repeats that localize to its Nterminal hydrophilic domain. The two known Akr1p substrates, the type I casein kinases Yck1p and Yck2p, presumably, are presented to Akr1p from the cytoplasm. When Yck palmitoylation is blocked, either through cis-mutation of their C-terminal, palmitoyl-accepting Cys-Cys dipeptides or through trans-mutation of the Akr1 PAT (i.e. akr1⌬), the kinases, rather than tethering to the plasma membrane, instead are left to distribute diffusely throughout the cytoplasm. To identify the Akr1p domains potentially available for substrate interaction and catalysis, we have determined, with the analysis below, a transmembrane topology for Akr1p.

Yeast Strains
The yeast strains used in this work (Table I) are isogenic to LRB759 (MAT␣ ura3-52 leu2 his3) (13) except at the AKR1 locus. The new strains were constructed by the previously described two-step gene replacement strategy (14).
Plasmids AKR1-SUC2-AKR1 Constructs-The starting point for the AKR1-SUC2-AKR1 constructs was a GAL1 P -AKR1(3xHA/FLAG/His) plasmid pND1436, which expresses from upstream GAL1 regulatory sequences, Akr1p C-terminally tagged by three iterations of the HA epitope, a single FLAG epitope, and then finally, at the C terminus, 6 histidine residues (1). The vector backbone for pND1436 is the yeast, single copy CEN/ARS/URA3 vector plasmid pRS316 (15). Insertion of the invertase-encoding SUC2 sequences into AKR1 utilized a strategy of in vivo gap repair in which the GAL1 P -AKR1(3xHA/FLAG/His) plasmid, linearized at the insertion site, was recircularized through in vivo homologous recombination with a chimeric DNA fragment consisting of the SUC2 sequences, flanked by 51 bp of upstream and downstream AKR1 homology. The first step in this process was to introduce, by oligonucleotide-directed mutagenesis (16), unique SmaI restriction sites at the desired insertion sites within the AKR1 coding sequence of pND1436. Ten sites, sites A-J (see Fig. 3A Insertion site J is located just C-terminally of the 3xHA/ FLAG/His tri-tag sequence. The AKR-SUC2-AKR1 chimeric DNA fragments for gap repair of the SmaI-linearized plasmids were generated by PCR. The upstream PCR primer consisted of the 51 bp of the AKR1 sequence immediately upstream of the insertion point followed at the 3Ј end by 19 bp of the SUC2 sequence that serves as the primer for the invertase segment. Likewise, the downstream primer had the 51 bp of the AKR1 sequence immediately downstream of the insertion point followed by the 19 bp of the SUC2 sequence, serving as the downstream primer of the invertase segment to be inserted. The PCR template was the SUC2-containing plasmid pSL883 (17). The PCR product together with the appropriate SmaI-linearized plasmid were co-introduced into LRB759 yeast cells, and Ura ϩ transformants were selected. Plasmids isolated from the yeast transformants were amplified in Escherichia coli, and the fidelity of the gap repair was assessed by restriction analysis. Two series of AKR1-SUC2-AKR1 constructs were made. The first series inserted a 46-residue-long invertase segment, invertase residues Gln 88 -Tyr 133 . This invertase segment, like the one used by Gilstrung and Ljungdahl (18), includes the three NX(S/T) sites at residues 97-99, 111-113, and 118 -120, which are known to be used for N-linked glycosylation in secreted invertase (19). The second series utilized the entire mature form of secreted invertase from Thr 32 (located just to the C-terminal side of the invertase signal peptide cleavage site) to the C-terminal invertase residue Lys 532 .
Other Plasmids-For the N-terminal AKR1 deletions (⌬2-21, ⌬2-70, and ⌬2-295), an upstream BamHI site, introduced by oligonucleotidedirected mutagenesis (16) between codons 1 and 2, was ligated to downstream BamHI sites, introduced at codons 21, 70, and 295. Similarly, the two C-terminal AKR1 deletions (⌬614 -764 and ⌬733-764) involved ligation of a downstream XhoI site, located just prior to the termination codon, to upstream XhoI sites that had been introduced at codons 614 and 733. The deletion constructs were carried on the pRS316 vector plasmid with expression controlled by upstream AKR1 regulatory sequences. For GAL1 P -STE2(3xHA), three HA epitope iterations replace the 6 C-terminal residues of the Ste2p C-terminal regulatory tail domain. For AKR1-GST(3xHA), the Schistosoma japonicum glutathione S-transferase (from pGEX-2T; Amersham Biosciences) is fused in-frame to the Akr1p C terminus, just prior to three iterations of the HA epitope.

Analysis of N-linked Glycosylation
Wild-type yeast cells (LRB759), transformed by plasmids for GAL1 Pdriven protein expression, were cultured overnight to log phase in YPR (1% yeast extract, 2% peptone, 2% raffinose). The addition of galactose to 2% initiated a 2-h period of protein expression, which was terminated by the subsequent addition of glucose to 3%. Following a final 20-min growth interval, 4 ϫ 10 7 cells were collected by centrifugation and resuspended in 0.18 ml of ice-cold lysis buffer (1 M sorbitol, 25 mM Tris/Cl, pH 8.0), trichloroacetic acid was added to 17%, and the cells were frozen at Ϫ80°C until the time of their processing to protein extracts. For some experiments, tunicamycin was added at 20 g/ml, 30 min prior to the addition of galactose.
To prepare protein extracts, the frozen cell suspensions were thawed on ice, and a 0.2-ml volume of 0.3-m glass beads (Sigma) was added followed by 10 min of vigorous vortexing. Two hundred and fifty microliters of 5% trichloroacetic acid were then added, and the cell lysate was decanted away from the rapidly settling beads into a fresh tube. The beads were subsequently washed with one 0.3-ml aliquot of 5% trichloroacetic acid, and the additional lysate recovered was combined with the initial lysate. Following 20 min on ice, the precipitated protein was collected by centrifugation (10 min, 13,000 rpm, 4°C). The protein pellet was washed with 1 ml of acetone at room temperature, and then, following desiccation, 100 l of sample buffer (5% SDS, 8 M urea, 40 mM Tris/Cl, pH 6.8, 0.1 mM EDTA) containing 1% ␤-mercaptoethanol was added, and the pellets were dissolved with a 10-min, 37°C incubation interspersed with vigorous vortexing. Six microliters of each sample was then subjected to SDS-PAGE, Western blotting, and finally, development with the anti-HA mAb conjugated with horseradish peroxidase (HRP) (Roche Applied Science).
This work a All strains are isogenic to LRB759 except at the AKR1 locus.

akr1⌬ Complementation
The different SUC2 insertion alleles and several N-and C-terminal AKR1 deletion alleles were tested for their ability to restore 37°C growth to akr1⌬ cells. The different mutant alleles were introduced into NDY1405 on the single copy URA3 yeast vector plasmid pRS316 under the expression control of natural AKR1 upstream regulatory sequences. Suspensions of the NDY1405 transformants were diluted and spottitered on uracil-lacking, minimal medium plates that were incubated either at 25 or at 37°C.

Biotinyl Replacement of Protein Acyl Modifications
A modification of the Drisdel and Green protocol (20), an in vitro exchange of protein acyl modifications for a biotinylated compound, was used to assess Yck2p palmitoylation in the different akr1 mutant cell contexts. Our modification is a proteomic adaptation that biotinylates and purifies palmitoylated proteins from complex protein extracts. 2 In brief, total protein prepared from the different akr1 mutant cells expressing N-terminally His/FLAG/HA tri-tagged Yck2p from the GAL1 promoter were subjected to the Drisdel and Green acyl-biotin exchange reactions (20). First, free thiols were blocked with N-ethylmaleimide; second, thioesters (i.e. cysteinyl-acyl linkages) were cleaved with 1 M hydroxylamine; and finally, thiols, newly exposed by the hydroxylamine treatment, were modified with a thiol-specific biotinylation reagent. Biotin-HPDP (Pierce), which disulfide-bonds to thiols, was substituted for the biotin-BMCC (Pierce) used by Drisdel and Green (20). Following biotinylation, epitope-tagged Yck2p was immune-precipitated with M2 anti-FLAG mAb-agarose and subjected to SDS-PAGE by Western analysis either with avidin-HRP (Pierce, Rockford, IL) to detect Yck2p biotinylation or with anti-HA-HRP to detect total recovery of the His/ FLAG/HA-tagged Yck2 protein.

Akr1p Multimerization Test
NDY1405 cells were doubly transformed by the AKR1-3xHA/FLAG/ His-CEN/ARS/LEU2 and the AKR1-GST/3xHA-CEN/ARS/URA3 plasmid constructs or by just one of the two AKR1 plasmids plus the complementarily marked empty vector plasmid, either pRS316 or pRS315 (15). The two plasmids were maintained in the cells during the course of the experiment by growth in minimal yeast medium, unsupplemented by uracil or leucine. Lysates were prepared under the native conditions previously used to purify active Akr1p (1). In brief, cells were broken by mortar and pestle grinding in liquid nitrogen, and then total yeast membranes were isolated and solubilized by Triton X-100. Finally, FLAG-tagged Akr1p was immune-precipitated with M2 anti-FLAG mAb-agarose (Sigma). Wash buffer was supplemented with 0.5 mg/ml bovine liver lipids (Avanti Polar Lipids, Alabaster, AL) to maintain bilayer-like conditions as described previously (1). Finally, purified proteins were eluted in sample buffer and subjected to anti-HA Western blotting.

RESULTS
Akr1p Hydropathy Analysis-Akr1p hydropathy analysis finds numerous sequences with hydrophobicity sufficient for service as TMDs (Fig. 1). In addition to the five strongly hydrophobic peaks, there are three sequences of marginal hydro-phobicity, at residues 182-201, 554 -568, and 661-680, that could potentially serve as TMDs, particularly within the context of a polytopic membrane protein. A further uncertainty relates to the broad bifurcated peak, corresponding to residues 324 -361, a 38-residue stretch of uncharged residues with the somewhat polar sequence SHINP, centrally embedded at residues 342-346. Does the 324 -361 correspond to one or two TMDs? Thus, sequence information alone predicts anywhere from 5-9 TMDs.
Analysis of Akr1-Suc2-Akr1 Constructs-As an experimental approach to the topology problem, we have opted to examine the usage of Asn-X-Ser/Thr (NX(S/T)) N-linked glycosylation sites introduced into the different Akr1p hydrophilic domains, the glycosylation sites being donated from segments of the secreted and heavily glycosylated yeast protein invertase (Suc2p). NX(S/T) sites introduced into the luminally oriented domains of Akr1p would be expected to be glycosylated, whereas those introduced into cytoplasmic domains should be unavailable to the luminally localized glycosyl transferases and thus should remain unglycosylated. Prior to embarking on this analysis, we first examined the usage of the four NX(S/T) sequences that are naturally present in the Akr1p sequence ( Fig. 1). A C-terminally, epitope-tagged Akr1p was expressed from the GAL1 promoter in the presence or absence of tunicamycin, a drug that blocks N-linked glycosylation (Fig. 2). The tagged AKR1 is apparently functional; the tagged AKR1 complements akr1⌬, restoring 37°C growth to the temperaturesensitive akr1⌬ cells (data not shown). As a control, we have also examined the glycosylation of the ␣-factor pheromone receptor Ste2p, a protein having three NX(S/T) sites that is known to receive N-linked glycosylation (21). Indeed, tunicamycin treatment does substantially alter gel mobility for Ste2p (Fig. 2). In contrast, Akr1p gel mobility is unaltered by tunicamycin, indicating that the four Akr1 NX(S/T) sites likely are not used for glycosylation.
To assess Akr1p topology, we made a series of Akr1-Suc2-Akr1 constructs that inserted a 46-residue-long segment of invertase, containing three NX(S/T) sites known to be utilized for glycosylation (19), at seven different positions within the AKR1 open reading frame (Fig. 3A, positions A, C, D, E, F, G, and H). This strategy had been successfully employed by Gilstring and Ljundahl (18) to assess topology of the yeast general amino acid permease Gap1p. Gel mobility of the seven Akr1-Suc2-Akr1 insertion proteins was compared (Fig. 3B). Our expectation, based on the published Gap1p analysis (18) 1. Akr1p hydropathy analysis. Upper panel, a hydropathy plot derived from the Akr1 protein sequence using TopPred II (27). Sequences exceeding the hydrophobicity threshold for service as TMDs are indicated in black. Lower panel, an Akr1p schematic with the strongly hydrophobic sequences indicated in black and with the more marginally hydrophobic sequences indicated in gray. The positions of the DHHC-CRD, the six ankyrin repeats, and the four potential NX(S/T) glycosylation sites (asterisks) are also indicated.

FIG. 2. Analysis of N-linked glycosylation for Akr1p and Ste2p.
Cells expressing HA epitope-tagged versions of Akr1p or of Ste2p were inducibly expressed from the GAL1 promoter. Thirty minutes prior to the galactose addition, cells either were treated with tunicamycin at 20 g/ml or were mock-treated. Following a 2-h expression period, protein extracts were prepared, and the extracts were subjected to SDS-PAGE and Western analysis with anti-HA mAb. In addition to being glycosylated, Ste2p is also subject to both phosphorylation and ubiquitination, which likely accounts for the residual heterogeneity of the tunicamycintreated protein.
inserted invertase sequences were luminally oriented. Surprisingly, no substantial mobility differences were seen among the seven Akr1-Suc2-Akr1 proteins; all migrated to essentially the same position, the position of the unglycosylated Akr1-Suc2-Akr1 protein (Fig. 3B). For Akr1-Suc2-Akr1-(D) and -(F), a minor subpopulation of the protein was found to diffusely migrate with retarded mobility (Fig. 3B). This rather subtle effect likely reflects a low level of heterogeneous glycosylation as this more slowly migrating protein smear is abolished with expression of the insertion proteins in tunicamycin-treated cells (data not shown). Nonetheless, this minor glycosylation effect is far more subtle than the striking glycosylation shifts seen for analogous Gap1-Suc2-Gap1 insertion constructs (18). Perhaps for the Akr1p structure, luminal segments are less well exposed and are thus less available to glycosyl transferases; indeed, glycosylation site usage is known to correlate with site availability during protein folding (22). Another possibility is that luminally oriented NX(S/T) sites of the Akr1-Suc2-Akr1 proteins are, in fact, glycosylated but simply with smaller, and thus less easily detectable, oligosaccharide moieties than those attached to Gap1p. Indeed, the elaboration of the N-linked core oligosaccharide varies tremendously for yeast glycoproteins; secreted proteins and proteins of the yeast cell wall generally have highly elaborated oligosaccharides with up to 100 mannose moieties added per oligosaccharide, whereas the glycosyl modifications of Golgi and vacuolar constituents generally are much less elaborate (23).
In hopes of magnifying the glycosylation effect, a second series of Akr1-Suc2-Akr1 constructs with virtually the entire invertase coding sequence inserted was generated. The 511residue-long invertase portion used corresponds to mature invertase in its entirety (minus the N-terminal signal peptide) with its 10 N-linked glycosylation sites. This invertase segment was inserted at 10 Akr1p sites: the seven positions utilized for the smaller, 46-residue Suc2 inserts plus three additional sites (Fig. 3A, sites B, I, and J). More striking gel mobility differences were seen for these second generation Akr1-Suc2-Akr1 proteins (Fig. 3C). Retarded mobilities consistent with glycosylation were evident for insertions at positions D, F, and H but not for insertions at positions A, B, C, E, G, I, and J. To test whether the up-shifted mobilities result from N-linked glycosylation, two of the Akr1-Suc2-Akr1 proteins were expressed in tunicamycin-treated cells. Indeed, tunicamycin does eliminate the Akr1-Suc2-Akr1-(H) up-shift, causing it to comigrate with the presumably unglycosylated Akr1-Suc2-Akr1-(E) (Fig. 3D).
We also note, for several of the insertion proteins, the presence of a secondary, minor gel band on the Western blot in addition to the major gel band (Fig. 3C). For instance, for Akr1-Suc2-Akr1-(D) and -(H), whereas the bulk of the protein runs at the glycosylated position, a minor species is also apparent for both, migrating at the position of non-glycosylated Akr1-Suc2-Akr1 protein. Similarly, for Akr1-Suc2-Akr1-(E), in which the bulk of the protein migrates to the unglycosylated position, an additional faint band is also apparent at the glycosylated position. This minor Akr1-Suc2-Akr1-(E) species is apparently glycosylated since it is both lost in tunicamycintreated cells (Fig. 3D) and also collapses to the unglycosylated position when extracts are treated prior to Western analysis with endoglycosidase H to remove N-linked oligosaccharides (data not shown). The existence of such minor species suggests that at least some of the insertion proteins can adopt alternative topologies in which domains that are normally cytoplasmic are instead oriented to the luminal compartment, or conversely, where luminal domains are instead oriented to the cytoplasm. Most likely, the 511-residue invertase segment inserted at these three Akr1 sites (i.e. at sites D, E, or H) is somewhat disruptive to the process by which Akr1p is assembled into the membrane, resulting in a low level of misinsertion. However, these minor electrophoretic species, even when most prominent, as for Akr1-Suc2-Akr1-(D) and -(H) insertions (Fig. 3C), represent less than 5% of the total Akr1-Suc2-Akr1 protein. Therefore, in developing an Akr1p topology model, we have concentrated on the glycosylation status of just the most prominent, major electrophoretic species, presumably representing the bulk topology for each insertion construct.
Akr1p Topology Model-Of the 10 Akr1-Suc2-Akr1 constructs examined, only three, Akr1-Suc2-Akr1-(D), -(F), and -(H), were found to be extensively glycosylated, indicating that just three sites (i.e. D, F, and H) are luminally oriented. The remaining seven constructs are not extensively glycosylated, indicating that the corresponding Akr1p sites (i.e. A, B, C, E, G, I, and J) all are cytoplasmic. A topology model derived from these results is shown in Fig. 4. Akr1p is found to have six TMDs with the bulk of the polypeptide being oriented to cytoplasm. Indeed, only three short connecting loop domains are  Fig. 1, strongly hydrophobic sequences are designated by black boxes, and more marginally hydrophobic sequences are designated by gray boxes. B and C, gel mobility assessed for Akr1-Suc2-Akr1 proteins with either the small, 46-residue invertase inserts (B) or the large, 511-residue invertase inserts (C). HA-tagged Akr1-Suc2-Akr1 proteins were expressed and detected by anti-HA Western blotting as described for Fig. 2. wt, wild type. D, retarded mobility of Akr1-Suc2-Akr1-(H)p is due to N-linked glycosylation. Cells were treated with (ϩ) or without (Ϫ) tunicamycin 30 min prior to the induction of Akr1p or Akr1-Suc2-Akr1p expression. Akr1-Suc2-Akr1 proteins have the large 511-residue invertase inserts. luminally oriented. The cytoplasmically oriented domains are: a large 323-residue-long N-terminal domain that includes the ankyrin repeats; two loop domains, one of which harbors the DHHC-CRD; and finally a 206-residue-long C-terminal domain. The various segments of marginal hydrophobicity noted previously (Fig. 1, gray boxes) are not TMDs, whereas the extended, 38-residue-long hydrophobic segment (residues 324 -361) corresponds apparently to two TMDs linked by an extremely short hydrophilic loop sequence.
Functionality of the Akr1-Suc2-Akr1 Proteins-To assess functionality, the different AKR1-SUC2-AKR1 alleles were tested for akr1⌬ complementation. Like many other of the yeast functions participating in endocytic and/or vacuolar trafficking, AKR1 gene deletion results in temperature sensitivity; thus, akr1⌬ cells, although viable at 30°C, are inviable at 37°C. This 37°C requirement for AKR1 appears to be a requirement for the Akr1 PAT function since akr1 missense alleles with substitutions of conserved DHHC-CRD residues both fail to support Yck2p palmitoylation in vivo and in vitro (1) and also fail to support 37°C viability (data not shown).
Complementation of the 10 AKR1-SUC2-AKR1 alleles with the large, 511-residue invertase insertion was tested first. akr1⌬ cells, harboring the different insertion alleles on single copy, centromeric plasmids, were plated either at 25 or at 37°C (Fig. 5A). Six of the constructs, namely insertions at sites A, D, F, H, I, and J, complemented, fully restoring 37°C growth to the akr1⌬ cells. The remaining four insertion alleles, B, C, E, and G, failed to restore 37°C growth. Interestingly, all four of these non-complementing insertions map to cytoplasmic domains; B and C map within the large N-terminal cytosolic domain, whereas E and G map to the two cytosolic loops. In contrast, insertions into the three luminal loop domains, i.e. sites D, F, and H, apparently do not disrupt Akr1p function (Fig. 5A). The functionality of the seven AKR1-SUC2-AKR1 alleles having the smaller, 46-residue invertase segment inserted were also examined (data not shown). Results with the smaller inserts were quite consistent with those for the larger inserts; A, D, E, F, and H complemented, whereas C and G failed to complement (data not shown). The one difference between the two sets of complementation data relates to insertions at position E, within one of the two cytosolic loops; although Akr1p function apparently is not abolished with insertion of the smaller 46-residue invertase segment (data not shown), function is lost with insertion of the larger invertase segment at the same site.
In addition, several N-and C-terminal AKR1 deletion alleles have been constructed and tested for complementation (Fig.  5B). Again, restoration of 37°C growth to akr1⌬ cells was assessed. The less severe deletions at both the N and the C termini restored growth, whereas the more severe ⌬2-295 Nterminal and ⌬615-764 C-terminal deletions failed to complement. We note for the non-complementing ⌬2-295 allele that the deletion extends into the ankyrin repeat sequence, whereas the deletions of the complementing N-terminal deletion alleles, ⌬2-21 and ⌬2-70, do not encroach upon this sequence. Together with the non-complementation of the AKR1-SUC2-AKR1-(B) allele (Fig. 5A), this suggests a likely requirement for the ankyrin repeats in Akr1p function.
PAT Function of Akr1-Suc2-Akr1 Proteins-Although we argue above that the restoration of 37°C growth to akr1⌬ cells likely represents a restoration of Akr1p PAT function, such a complementation assay is obviously quite an indirect measure of PAT function. As a more direct assessment of PAT function, we have examined several AKR1-SUC2-AKR1 insertion alleles for their support of Yck2p palmitoylation in vivo. Rather than using the standard method for assessing palmitoylation, i.e. examining the in vivo incorporation of label from [ 3 H]palmitic acid into the protein of interest (Yck2p in this case), we have opted to use instead a recently published protocol involving a chemical exchange of the protein acyl modifications for an introduced labeling compound (20). For this, protein is subjected to three sequential chemical treatments. First, free thiols are exhaustively blocked with N-ethyl-maleimide. Second, palmitoyl moieties are released from modified cysteines through specific hydroxylamine cleavage of the thioester linkage. Finally, the thiols newly exposed by the hydroxylamine treatment (i.e. the cysteine residues that were palmitoylated) are marked with a labeled thiol reagent. This method is more robust than the in vivo palmitate labeling approach and quite specific for detection of palmitoyl modifications (20). Yck2p palmitoylation was examined in strains utilizing three of the different AKR1-SUC2-AKR1 insertion alleles (511-residue-long Suc2 insertions) in place of chromosomal AKR1 (Fig. 6). As a control, palmitoylation of a mutant Yck2p with its C-terminal palmitoyl-accepting Cys-Cys sequence mutated to Ser-Ser, i.e.  (15). Equal aliquots of each transformant were spot-titered on minimal yeast plates lacking uracil which were subsequently incubated at either 24 or 37°C. The growth of the 10 AKR1-SUC2-AKR1 transformants was compared with akr1⌬ cells transformed either by the pRS316 empty vector (Ϫ) or by the pRS316borne wild-type AKR1 allele (wt). B, AKR1 N-and C-terminal deletions. The indicated AKR1 deletion alleles were introduced into akr1⌬ cells and tested for their ability to restore 37°C growth as in panel A.
Yck2(SS)p, also was tested. Consistent with previous analyses of Yck2p palmitoylation by standard in vivo [ 3 H]palmitate labeling (1,24), the new methodology finds Yck2p to be palmitoylated when isolated from wild-type cells but not when isolated from akr1⌬ cells (Fig. 6); furthermore, the mutant Yck2(SS)p, which lacks the palmitoyl-accepting cysteines, also failed to get biotinylated (Fig. 6). Of the three AKR1-SUC2-AKR1 alleles tested, insertions at sites D and F supported Yck2p palmitoylation, whereas the G insertion allele did not. These results are fully consistent with the complementation analysis; the D and F insertion alleles restored 37°C growth to the akr1⌬ cells, whereas the G insertion allele did not (Fig. 5A). Thus, for D and F, and likely also for A, H, I and J, the invertase insertions do not abolish Yck2p PAT function.
In addition, we note a striking effect on Yck2p gel mobility shift that correlates with palmitoylation status (Fig. 6, lower  panel). In the three contexts in which Yck2p fails to be palmitoylated, i.e. for Yck2(SS)p and for wild-type Yck2p in either akr1⌬ or AKR1-SUC2-AKR1(G) cells, Yck2p runs as a single coherent gel band. However, when palmitoylated, an additional, more prominent, slower migrating species is also seen. Rather than reflecting a direct effect of the palmitoyl modification on gel mobility, the more slowly migrating species is instead a consequence of phosphorylation; following phosphatase treatment, palmitoylated and non-palmitoylated Yck2 proteins comigrate (1) (data not shown). Indeed, we have found that the Yck2p hyperphosphorylation is compartment-specific, depending on Yck2p transport to the plasma membrane. 3 Akr1p Multimerization Test-In our published analysis of the Akr1p-associated PAT activity, Akr1p was purified to apparent homogeneity; we found no evidence for other stoichiometric components that co-purified (1). Thus, unlike Erf2p and Erf4p, which do co-purify and which appear to behave as cosubunits for Ras palmitoylation, for Akr1p, no accessory proteins appear to be required. Nonetheless, the active PAT could still be composed of multiple copies of the Akr1 polypeptide. To test for such homo-multimerization, we co-expressed two Akr1 proteins, differing in the sequences fused at their C termini. Akr1-3xHA/FLAG/His has the tri-tag sequence previously used in the affinity purification of the active Akr PAT (1). Akr-GST/3xHA has at its C terminus the 223-residue bacterial glutathione S-transferase (GST) followed by three iterations of the HA epitope tag. Both constructs fully complement akr1⌬, restoring 37°C growth (data not shown), and the two proteins can be easily differentiated by their quite different gel mobilities (Fig. 7). Lysates, prepared from akr1⌬ cells expressing the two proteins either separately or together, were subjected to anti-FLAG affinity purification under the native conditions previously used to purify the Akr1 PAT activity (1). The Akr1-3xHA/FLAG/His protein is efficiently purified by the anti-FLAG immune precipitation, yet we find no evidence for copurification of the co-expressed Akr1-GST/3xHA protein (Fig.  7). Thus, we find no evidence for Akr1p homo-multimerization. DISCUSSION Our results indicate that the Akr1 polypeptide chain crosses the bilayer six times with its major hydrophilic domains oriented to the cytoplasm (Fig. 4). Cytoplasmic domains include the ankyrin repeat-containing, 323-residue-long N-terminal domain, the 166-residue-long C-terminal domain, and two inter-TMD loop domains, one of which contains the DHHC-CRD sequence. The luminally exposed domains consist of three relatively short loop segments. Such a topology, in which the bulk of the protein is accessible from the cytoplasm, fits well with the known biology for Akr1p. Its two known substrates, the kinases Yck1p and Yck2p, are expected to be presented to Akr1p from the cytoplasm; Yck1p and Yck2p are synthesized initially as soluble, cytoplasmic proteins, and indeed, when their palmitoylation is blocked, they are found to diffusely distribute through the cytoplasm (1,25). The acyl donor for the palmitoylation reaction, palmitoyl-CoA, given its amphiphilic nature, might be expected to dissolve in membranes and thus gain access to the PAT through the bilayer. However, it has been argued that the overwhelming bulk of the acyl-CoA pool is sequestered within the cytoplasm by the abundant cytoplasmic protein, acyl-CoA-binding protein (26), or Acb1p in yeast. By shielding the acyl-CoA hydrophobic acyl tail from the aqueous environment, acyl-CoA-binding protein is thought to potentiate acyl-CoA transport through the cytoplasm. Thus, like the substrate proteins, palmitoyl-CoA may also access the PAT from the cytoplasm. Finally, the finding that the Akr1p DHHC-CRD sequence is cytoplasmically oriented suggests that 3 6. PAT function of AKR1-SUC2-AKR1 alleles. Yeast strains with AKR1-SUC2-AKR1-(D), -(F), or -(G) alleles in place of chromosomal AKR1 were compared with both wild-type cells (wt) and akr1⌬ cells (⌬) for their support of Yck2p palmitoylation. As an additional control, the palmitoylation of Yck2(SS)p, a mutant having Ser-Ser substituting for the C-terminal, palmitoyl-accepting Cys-Cys dipeptide, was also examined. Cell culturing and the galactose induction of epitope-tagged Yck2p expression from a GAL1 P -His/FLAG/HA-YCK2 plasmid (1) were as described in the legend for Fig. 2. Cell lysates were prepared and subjected to the acyl-biotin exchange reactions (see "Experimental Procedures"). The tri-tagged Yck2p, immune-precipitated from the biotinylated protein extracts with anti-FLAG mAb-agarose, was subjected to Western analysis both with avidin-HRP to detect biotinylation and with anti-HA-HRP to assess Yck2p recovery.
FIG. 7. Akr1p multimerization test. NDY1405 yeast cells were doubly transformed by two centromeric plasmids, each expressing one of two functional but electrophoretically distinct versions of Akr1p: Akr1-3xHA/FLAG/His (1) having the 54-residue tri-tag sequence at the Akr1p C terminus (Akr1-FLAG) or Akr1-GST/3xHA, which has 223residue bacterial glutathione S-transferase followed by three repeats of the HA epitope at the Akr1p C terminus (Akr1-GST). Lysates, prepared under the native conditions previously used to maintain Akr1 PAT activity (1), were subjected to immune precipitation (IP) with anti-FLAG mAb-agarose followed by anti-HA Western blotting. As controls, cells singly transformed by the Akr1-3xHA/FLAG/His-expressing and Akr1-GST/3xHA-expressing plasmid were processed in parallel. catalysis, i.e. the addition of the acyl moiety to the target cysteine residue, may also be mediated by cytoplasmic domains. The role for the DHHC-CRD in catalysis was suggested by previous findings made for both Akr1p and Erf2p that mutation of conserved DHHC-CRD residues abolished PAT activity in vivo and in vitro (1,2). In addition to fitting well with the biology of Akr1p, the presented topology, with the bulk of the PAT polypeptide oriented to the cytoplasm, also more generally fits with palmitoylation as a cytoplasmically limited protein modification (6 -8).
A central role for the cytoplasmic Akr1p domains is also supported by the analysis of the function of the different Akr1-Suc2-Akr1 insertions. These mutations insert either 46 or 511 residues at different sites within the Akr1 polypeptide and thus have great potential for disrupting Akr1 function. Although some insertions do abolish Akr1p function, surprisingly, many do not. Indeed, in addition to withstanding fusions to both the N and the C termini, insertions at multiple internal sites also appear not to disrupt Akr1p function. Perhaps most significantly, all of three luminally oriented insertions, i.e. the insertions into the three luminally oriented loop domains of Akr1p, remain grossly functional. In fact, two of the three luminal insertions were tested for their in vivo support of Yck2p palmitoylation and were found to be fully functional; no loss of Akr1 PAT function was detected. The four insertions that were found to disrupt function all mapped to cytoplasmic domains (sites B, C, E, and G). At site E, within the more N-terminal of the two cytoplasmic loop domains, different results were found for the small versus the large Suc2 insertions; function was disrupted by the large, but not by the small, Suc2 insertion. Thus, by this crude mutational analysis, the Akr1p cytoplasmic domains appear to play a more critical role in Akr1p function than do the luminal domains.
How does the Akr1p topology model (Fig. 4) fit with the hydropathy analysis ( Fig. 1)? First, the three segments of marginal hydrophobicity, residues 182-201, 554 -568, and 661-680 ( Fig. 1, gray boxes of the Akr1p schematic), apparently do not serve as TMDs. Secondly, the relatively broad, 38-residue-long 324 -361 hydrophobic sequence harbors two TMDs. Central within this extended hydrophobic segment is a short hydrophilic subsegment, the pentapeptide sequence SHINP (residues 342-346), which apparently constitutes an extremely short, luminally disposed, inter-TMD loop domain. Invertase inserted into this pentapeptide sequence (Akr1-Suc2-Akr1-(D)) gets glycosylated, whereas invertase inserted at the C and E flanking sites does not (Fig. 3, A and C). Although it seems unusual for two TMDs to be so closely spaced within the sequence, we note that the yeast and human Akr1p homologues (for which homology extends across the entire protein sequence) have preserved similarly positioned hydrophobic domains (Fig. 8). At the homologous site within yeast Akr2p, there is a contiguous stretch of 43 uncharged residues, i.e. residues 298 -340 (Fig. 8). The uncharged Akr2p sequence stretch is divided by a central SLVLSP (residues 317-322), which is somewhat less hydrophilic than the corresponding SHINP loop segment of Akr1p. The two human Akr1p orthologues, HIP14 and HIP14L, have, at the equivalent position, two discrete hydrophobic sequences, clearly delineated by an intervening hydrophilic sequence with multiple charged residues rather than the single long hydrophobic sequence found in Akr1p and Akr2p. It is perhaps noteworthy that for each of the four proteins, the second TMD of this pair is notably short, consisting of 15 uncharged residues in Akr1p, 18 residues for Akr2p, and 15 residues for both HIP14 and HIP14L.
In addition to the three Akr1p homologues discussed above (i.e. Akr2p, HIP14, and HIP14L), Fig. 8 also compares the Akr1p hydropathy with hydropathies for other DHHC proteins: the other five yeast DHHC proteins as well as rat GODZ, which has recently been implicated in the palmitoylation of the ␥2 subunit of the ionotropic GABA A receptor (12). As homology within this family is largely limited to the DHHC-CRD, we have used this conserved domain as a reference point for plot alignment. To predict topologies for each (Fig. 8, bars above each hydropathy plot), we made the assumption that the DHHC-CRD is oriented cytoplasmically as it is for Akr1p. Clear similarities are evident across the diverse group with regard to the spacing of TMDs that surround the DHHC-CRD. Most FIG. 8. Hydropathy analysis of yeast and mammalian DHHC proteins. Hydropathy plots of nine DHHC proteins are compared with the Akr1p plot. Included are the plots for all seven yeast DHHC proteins as well as for three selected mammalian DHHC proteins, the two human orthologues of Akr1p, namely HIP14 and HIP14L, and the rat DHHC protein GODZ. Plots, derived as described in the legend for Fig.  1, are aligned via their conserved DHHC-CRD. Positions of the DHHC-CRD (blue bar) and ankyrin repeats (red bar) are indicated. Based on the assumption that the DHHC-CRD is cytoplasmically oriented, a prediction of topology is shown for each above the hydropathy plot. The orientation of predicted domains is indicated: TMDs (black), luminal domains (gray), and cytoplasmic domains (white). A core region with preserved overall hydropathy, consisting of four TMDs surrounding the DHHC-CRD, is indicated by the broad gold bar that extends through all 10 plots. notable is the fixed spacing of the DHHC-CRD to the C-terminally adjacent TMD, the spacing being constrained by the overlap of the DHHC-CRD homology with this TMD, which generally involves the 7 C-terminal DHHC-CRD residues. This invariant overlap of the putative active site (the DHHC-CRD) and the TMD suggests the possibility that essential aspects of the catalysis may occur within the bilayer plane. Extending further away from the DHHC-CRD, additional similarities are evident; the aligned plots show a roughly preserved, 220-residue core consisting of the four TMDs that surround the central DHHC-CRD (Fig. 8, gold bar extending through all of the plots). We will be interested to see whether this preserved portion represents a preserved PAT structural core for the DHHC protein family.